Hobbyist training AI on Victorian texts gets an unexpected history lesson from his own creation.

Ars Technica is known for its  coverage of technology-related news and analysis, ranging from scientific breakthroughs to the latest gadgets and gaming developments. Readers can learn about emerging technologies, industry trends, and the societal impact of technological advancements through detailed articles and reviews.

Ars Technica

A computer science student created TimeCapsuleLLM, an AI language model trained exclusively on Victorian-era London texts from 1800-1875. When prompted with "It was the year of our Lord 1834," the model unexpectedly generated text referencing real historical protests and Lord Palmerston's actions from that exact year. The student discovered through fact-checking that these were actual historical events, demonstrating how AI models trained on period texts can inadvertently capture and reproduce authentic historical information. This project is part of a growing field of Historical Large Language Models (HLLMs) that aim to recreate linguistic patterns and knowledge frameworks from past eras.

College student’s “time travel” AI experiment accidentally outputs real 1834 history

How The Callisto Protocol's Gameplay Was Perfected Months Before Release

“Trained on Vicotrian-era London texts from 1800-1875”
What do you mean unexpectedly generate text referencing the data it was trained on?

So it was trained on “Vicotrian-era London texts from 1800-1875” then gives Vicotrian-era London 1800-1875 history and the student was then surprised ?

Sounds like an llm wrote an article about an llm that the llm thought humans would like.

“an AI language model trained exclusively on Victorian-era London texts from 1800-1875”
“When prompted with “It was the year of our Lord 1834,” the model unexpectedly generated text …”
Hows that unexpected exactly? EXCLUSIVELY trained on that data btw.